After years of delays and hype, Intel has finally launched its first real 64bit processor, the Itanium 2. Now that the chip has been launched, the question is how long will it take for the rest of the industry to catch up.
The first crop of systems are unlikely to challenge the capabilities of existing high-end platforms. For example, server vendors that can't offer partitioning tools to improve the manageability of large systems will find it difficult to sell anything bigger than four- or eight-way configurations.
It might also take some time to convince anyone that the Itanium 2 provides reasonable performance, given that the Pentium 4 is much cheaper and beats it in simple integer benchmarks.
In the longer term, the Itanium 2's reliance on an optimising compiler may turn out to be the processor's Achilles' heel.
Intel is quick to point out the advantages of an optimising compiler. It says rather than dedicate some of the processor's hardware to find the best way to handle software, it gets the compiler to organise things so software makes the best use of the chip's resources.
The end result is that software made using the appropriate compiler runs at maximum speed, and other programs don't.
In effect, this means Itanium's Epic architecture actually forces independent software vendors to remain in lock-step with Intel's Itanium developers - a group of people not known for their ability to hit deadlines.
So software vendors will have to wait until Itanium 2 servers are brought to market before they can start the final beta tests and then release the final versions of their software, which all takes time.
Assuming about two months for each beta cycle, Intel will be lucky to see a shrink-wrapped version of Windows for the Itanium until close to Christmas. It follows that application vendors such as Oracle won't be able to deliver their Itanium 2 products until a little while after that.
But the biggest problem is that the software business must go through the same cycle of delay each time Intel updates the number of processing units inside the chip. This is scheduled to happen again about two years from now, and I wouldn't be surprised if further tweaks are also planned for the future.
So the Itanium's reliance on an optimising compiler also means we need a different attitude from software vendors about how they withdraw support for old versions of software, because if they stop their support too early it could force users to stop using their Itanium servers before they are ready to do so.
Of course, it is highly unlikely that software vendors will be able to afford to maintain a special version of their software for each Itanium architecture.
Given that microprocessors generally have a lifespan of around 10 years, this means the Itanium will be approaching the end of its life before much of the software industry has come to grips with this brave new chip.
Have your say: contact IT Week
reader comments